Journal of Family Psychology

© 2018 The Author(s) 2019, Vol. 33, No. 2, 215–225

0893-3200/19/$12.00 http://dx.doi.org/10.1037/fam0000492

Coparenting and Parenting Pathways From the Couple Relationship to

Children’s Behavior Problems
Alison Parkes, Michael Green, and Kirstin Mitchell
University of Glasgow

Although an extensive literature has linked couple conflict with the development of children’s externalizing
behavior problems, longer term protective effects of positive dimensions of couple relationships on children’s
externalizing behavior remain understudied, particularly in relation to underlying mechanisms. Supportiveness
in the dyadic couple relationship may enhance mothers’ and fathers’ individual parenting skills and protect
against children’s behavior problems, but the contribution of coparenting (couples’ support for one another’s
individual parenting) remains unclear. This observational study investigated associations between couple
supportiveness in children’s infancy and middle childhood externalizing problems, exploring pathways
involving coparenting and/or mothers’ and fathers’ individual parenting using data from the U.K. Millennium
Cohort Study (MCS; N ⫽ 5,779) and the U.S. Fragile Families and Child Wellbeing Study (FFS; N ⫽ 2,069).
Couple supportiveness was associated with reduced externalizing problems 8 to 10 years later (standardized
betas: MCS ⫽ –.13, FFS ⫽ ⫺.11, both ps ⬍ .001). Much of this effect (60% MCS, 55% FFS) was attributable
to coparenting and parenting when children were aged 3 to 5 years. Pathways from couple supportiveness
involving negative parenting were stronger than those via positive parenting, pathways via mothers’ parenting
were stronger than those via fathers’ parenting, and there were pathways via coparenting alone (without
affecting parenting). Pathways involving coparenting were similar in magnitude (MCS), or larger (FFS), than
those involving parenting alone. Consistent findings across different population samples suggest that helping
parents to support one another in coparenting and to develop their individual parenting skills may lessen the
longer term impact of couple relationship problems during early childhood.

Keywords: couple relationship, parenting, coparenting, externalizing behavior problems

Supplemental materials: http://dx.doi.org/10.1037/fam0000492.supp

Externalizing problems such as aggression, rule-breaking, and adverse adult outcomes (Odgers et al., 2008). Early childhood is a
attentional problems often develop early in childhood (0 to 5 key period of social and cognitive development that can protect
years) and then typically decline across middle childhood (6 to 12 against later externalizing problems (Fearon, Bakermans-
years; Petersen, Bates, Dodge, Lansford, & Pettit, 2015), but Kranenburg, van Ijzendoorn, Lapsley, & Roisman, 2010; Schoe-
persisting or increasing problems during middle childhood signal maker, Mulder, Deković, & Matthys, 2013), so it is critical to
compromised psychosocial functioning across a wide range of understand potentially modifiable aspects of family functioning
that can impinge on this early development. Yet despite the theo-
rized central importance of the couple relationship in supporting
This article was published Online First December 27, 2018. the key role of parenting during the child’s early years (Belsky,
Alison Parkes, Michael Green, and Kirstin Mitchell, MRC/CSO Social 1984), surprisingly little is known about how the couple relation-
and Public Health Sciences Unit, College of Medical, Veterinary & Life ship during early childhood affects children’s longer-term devel-
Sciences, University of Glasgow. opment (Pendry & Adam, 2013). Two meta-analyses dating from
This research was supported by Medical Research Council Grants the 1990s established that negative aspects of couple relationship
MC_UU_12017/11 and MC_UU_12017/13 and Chief Scientist Office
quality, especially conflict, constitute a small to moderate risk
Grants SPHSU11 and SPHSU13.
This article has been published under the terms of the Creative Com-
factor for children’s externalizing behavior (Buehler et al., 1997;
mons Attribution License (http://creativecommons.org/licenses/by/3.0/), Reid & Crisafulli, 1990). Conversely, positive aspects of the
which permits unrestricted use, distribution, and reproduction in any me- couple relationship such as satisfaction and supportiveness, while
dium, provided the original author and source are credited. Copyright for less well-studied than conflict, have protective effects (e.g., Gold-
this article is retained by the author(s). Author(s) grant(s) the American berg & Carlson, 2014). Most existing longitudinal studies are,
Psychological Association the exclusive right to publish the article and however, limited to relatively short follow-up periods, although a
identify itself as the original publisher.
recent study found that interparental conflict in early childhood
Correspondence concerning this article should be addressed to Alison
Parkes, MRC/CSOSocial and Public Health Sciences Unit, College of
predicted higher levels of externalizing problems from age 5 to
Medical, Veterinary & Life Sciences, University of Glasgow, Top Floor, adulthood (Petersen et al., 2015).
200, Renfield Street, Glasgow G2 3QB, United Kingdom. E-mail: More research is required to establish whether there are similar
[email protected] long term effects of positive dimensions of couple relationships,

215
216 PARKES, GREEN, AND MITCHELL

and to explore underlying mechanisms. This study explores asso- well-established, as many studies have linked less positive parent-
ciations between couple supportiveness in the child’s infancy and ing (e.g., warmth) and more negative parenting (e.g., harsh con-
middle childhood externalizing behavior problems. It investigates trol) with the development of externalizing problem behavior
possible pathways, via coparenting and parenting. (Pinquart, 2017). The model also suggests a pathway involving
coparenting alone, since there is empirical evidence for indepen-
Indirect Effects of Couple Relationship Quality via dent effects of coparenting and parenting on children’s external-
Parenting and Coparenting izing behavior (Teubert & Pinquart, 2010). This pathway is less
well understood than those involving parenting. Overt coparenting
The main theory indirectly linking the quality of the dyadic disagreements might model irritability, anger and aggression, dis-
couple relationship with child socioemotional adjustment, the inhibiting the use of such behavior regardless of parents’ individ-
spillover hypothesis, suggests two main mechanisms involving ual parenting effectiveness (Chen & Johnston, 2012). Even when
disruptions to family processes concerning the child (Erel & Bur- the child is not directly exposed to coparenting tensions, lack of
man, 1995). First, tensions in the couple relationship may lead to
cooperation and/or undermining coparenting might model negative
compromised parent– child interactions (parenting), as parents be-
behavior strategies, such as noncompliance with rules or bullying
come less emotionally sensitive to the child’s needs. This idea
(Murphy, Jacobvitz, & Hazen, 2016).
receives considerable empirical support, in relation to both couple
There is mounting empirical evidence for long-term indirect
conflict and lower marital satisfaction (Erel & Burman, 1995;
effects of the couple relationship in early childhood on child
Krishnakumar & Buehler, 2000). Second, couple relationship
behavior, via enduring effects on parenting quality. Several longi-
problems may undermine supportive coparenting. Coparenting
tudinal studies of families with young children have found indirect
concerns parents’ joint family management and division of labor,
paths from the couple relationship via positive and negative par-
agreement on child rearing and support for one another’s parenting
enting, acting over one or more years (Gerard, Krishnakumar, &
(Feinberg, 2003). Although often operationalised as a particular
domain of dyadic couple functioning, a family systems approach Buehler, 2006; Lindsey, Caldera, & Tankersley, 2009; Rhoades et
classifies coparenting as triadic functioning, where the child is also al., 2011; Schoppe-Sullivan, Schermerhorn, & Cummings, 2007;
involved (Talbot & McHale, 2004). Empirical work supports a Stover et al., 2016). In contrast, research that explores coparenting
distinction between triadic coparenting and dyadic couple relation- as an additional mediator is relatively sparse: two studies (O’Leary
ship functioning, finding that some couples with a poor dyadic & Vidair, 2005; Stroud, Meyers, Wilson, & Durbin, 2015) con-
relationship may engage in supportive triadic coparenting, and vice sider both coparenting and parenting as mediators, finding indirect
versa (Feinberg, 2003). Nonetheless, positive qualities of the dy- pathways via both types of process to externalizing problems.
adic couple relationship are likely to facilitate supportive copar- However, only one study allowed for all the paths suggested by our
enting, which may in turn help sustain dyadic parent– child rela- conceptual model (O’Leary & Vidair, 2005), and neither used a
tionships (Fincham & Hall, 2005). Empirical support for the idea longitudinal design.
that coparenting is an intermediary on spillover paths from the In addition to a paucity of research on a more comprehensive set
dyadic couple relationship to mothers’ and fathers’ individual of indirect pathways, there are other notable research gaps left by
parenting comes from several studies of families with young existing pathway studies of families with young children. The
children, (e.g., Holland & McElwain, 2013; Pedro, Ribeiro, & studies cited above examining pathways from the couple relation-
Shelton, 2012). ship to children’s externalizing problems via parenting all used
The conceptual model in Figure 1 extends the spillover model of couple conflict or hostility as the main exposure, and there is a
couple relationship qualities onto coparenting and parenting (Erel need to explore pathways to children’s problems from alternative,
& Burman, 1995; Fincham & Hall, 2005), to show their impact on more positive dimensions of the couple relationship. There are also
children’s externalizing behavior. It suggests indirect paths from question marks over the generalizability of existing studies. All but
the couple relationship via parenting alone, via coparenting and one (Gerard et al., 2006) of the parenting pathway studies relied on
then parenting, and via coparenting alone. The part of the model small, convenience samples with little or no allowance for poten-
extending parenting spillover to impact externalizing problems is tial confounders. A further limitation of existing longitudinal path-

Mother’s
parenting

Couple Child externalizing

Co-parenting
supportiveness behavior

Father’s
parenting

Figure 1. Conceptual model of pathways from couple relationship quality to children’s externalizing problems.
 COUPLE SUPPORTIVENESS AND CHILD BEHAVIOR PROBLEMS 217

way studies is that they do not generally span the critical period of Ward George, 2010). Empirical studies addressing the question of
the child’s transition to school, relating either to preschool (Lind- whether spillover affects one parent more than the other have
sey et al., 2009; Rhoades et al., 2011) or to school-age years inconsistent findings (Lindsey et al., 2009; Rhoades et al., 2011;
(Gerard et al., 2006; Schoppe-Sullivan et al., 2007). The one Stover et al., 2016; Stroud et al., 2015). In connection with our
exception covering school transition found that effects of the second hypothesis, we explore the following comparisons related
couple relationship were confined to the preschool period (Stover to the magnitude of distinct indirect pathways to clarify which
et al., 2016). More studies are therefore required to establish mechanisms are more dominant:
whether indirect pathways extend from couple relationship quality
in early childhood, to impact on school-age externalizing prob- (1) coparenting (with/without parenting) versus parenting alone,
lems.
(2) negative parenting versus positive parenting, and

Study Aims and Hypotheses (3) mother’s parenting versus father’s parenting.
This is the first longitudinal study to assess the independent We use two large longitudinal population-based samples, drawn
contribution of coparenting, as well as parenting, toward explain- from different countries, with different ethnic and sociodemographic
ing the effects of couple relationship quality during children’s compositions, so similar results will increase our confidence in the
early years on middle childhood externalizing problems. We de- generalizability of the mechanisms. The two study samples have
fine relationship quality in terms of overall mutual emotional distinct, complementary characteristics. The first consists of a sample
supportiveness, as this is likely to convey important positive as of stable couple families containing both biological parents coresiding
well as negative dimensions of couple functioning (Goldberg & throughout the child’s early years. As a nationally representative
Carlson, 2014). We assess the couple relationship in infancy as this sample, the U.K. Millennium Cohort Study (MCS) offers more gen-
is a good guide to subsequent supportiveness across the preschool eralizability than existing studies using convenience samples. The
years (Howard & Brooks-Gunn, 2009), and predates the develop- second draws from a large population-based study from the United
ment of behavior problems thereby excluding reverse causation States, the Fragile Families and Child Wellbeing Study (FFS). This
(but not confounding). We assess potential coparenting and par- study was designed to oversample unmarried parents, who are under-
enting mediators mainly in the preschool years (including some represented in research on couple relationship problems in relation to
information from the school transition period in one of the study child adjustment. Unlike the U.K. study, the U.S. study included data
samples). Coparenting and parenting of preschoolers is important on couple supportiveness, coparenting, and parenting from non-
for the development of executive functioning and secure attach- coresiding couples.
ment (Fay-Stammbach, Hawes, & Meredith, 2014; Karreman, van
Tuijl, van Aken, & Dekovic, 2008; Moss, Cyr, Bureau, Tarabulsy, Method
& Dubois-Comtois, 2005; Pérez, Moessner, & Santelices, 2017),
which in turn protect against externalizing problem development This study drew on data from two cohort studies. As a secondary
(Fearon et al., 2010; Schoemaker et al., 2013). Thus, we are able data analysis, ethical approval was deemed unnecessary.
to study our potential mediators at an important point. Nonethe-
Sample 1: MCS
less, because coparenting and parenting behaviors are moderately
stable in early childhood (Dallaire & Weinraub, 2005; Laxman et The MCS is a prospective study of U.K. children born between
al., 2013), we are likely to also capture the quality of these September 2000 and January 2002. The stratified clustered sam-
processes during the child’s infancy and toddler years. pling design ensures good representation of children from disad-
We explore the following hypotheses: (1) couple supportiveness vantaged areas, ethnic minority groups and from all four U.K.
will be associated with reduced childhood externalizing problems countries (Plewis, Calderwood, Hawkes, Hughes, & Joshi, 2007).
in middle childhood and (2) associations will be mediated by Families were first interviewed in-home at 9 months postbirth,
supportive coparenting and higher quality mother’s and father’s when 18,552 families were contacted (96% of those eligible to take
individual parenting of the preschool child. On the basis of previ- part). Interviews were repeated when children were aged 3, 5, 7,
ous literature, finding independent effects of parenting and copa- and 11, and teachers completed postal surveys at ages 7 and 11.
renting on externalizing problems (Teubert & Pinquart, 2010), we We excluded families with respondents in Scotland and Northern
expect additive indirect effects of coparenting and parenting con- Ireland (because teachers were surveyed only in England and
structs, although it is not clear whether coparenting (followed by Wales; n ⫽ 4,259), families in England and Wales with multiple
parenting or alone) channels most of the spillover effect of the births (n ⫽ 194), and those where the household did not contain
couple relationship (O’Leary & Vidair, 2005; Stroud et al., 2015). two resident parents (as only coresident parents were interviewed;
Previous pathway studies also lead us to expect indirect effects n ⫽ 2,471). Out of the remaining 11,628 couple families from
involving both negative and positive parenting (Gerard et al., England and Wales with a singleton child, 8,721 families were
2006; Lindsey et al., 2009; Schoppe-Sullivan et al., 2007; Stover et followed up at ages 3 and 5. Because full information on mediator
al., 2016), but it is not clear where the balance of effects lies. measures was only collected from coresiding couples living to-
Another unresolved issue is whether spillover has a more pro- gether at these time points, we dropped families where couples had
nounced effect on one parent; either the mother as the child’s main separated (n ⫽ 1,276), leaving 7,445 eligible families. The analytic
caregiver, or the father, whose involvement with the child, as sample was restricted to 5,779 of these, where information on
suggested by the fathering vulnerability hypothesis, may be rela- externalizing problems at the final time point, age 11, was pro-
tively dependent on partner support (Cummings, Merrilees, & vided by at least one source (parent, teacher, child).
218 PARKES, GREEN, AND MITCHELL

Sample 2: FFS items concerning the extent to which the partner compromised,
listened, empathized, offered encouragement, or expressed affec-
The FFS is a longitudinal study of families with a child born tion.
between 1998 and 2000 in 20 large U.S. cities (⬎200,000 inhab- To reflect the dyadic nature of the couple relationship construct,
itants; N ⫽ 4,897). It was designed to oversample unmarried an average of the two parents’ scores was used in the main
couples: these represented three quarters of the sample at baseline, analysis. Supplementary analyses address the potential relevance
with 40% not living together when their child was born (Reich- of different perceptions of the couple relationship by each parent,
man, Teitler, Garfinkel, & McLanahan, 2001). Mothers and fathers which was particularly important given relatively low correlations
were interviewed in the hospital within 48 hr of the child’s birth, between ratings of the couple relationship in FFS (r ⫽ .19, p ⬍
and follow-up information was collected by telephone when the .001, compared with r ⫽ .44, p ⬍ .001 in MCS).
child was approximately 1, 3, 5, and 9 years old. In-home assess- Potential mediators. Both parents reported on coparenting
ments were performed at 3, 5, and 9 years, and teachers completed and parenting when children were aged 3 (or age 5 for some
postal questionnaires at 5 and 9 years. At child aged 1 year, there measures in MCS), described as follows.
were 4,364 families where the mother was interviewed. We ex- Coparenting. For MCS, this was based on the average score
cluded families with multiple births (n ⫽ 83), families where the of one item reported at child aged 3 and 5 by each parent on how
mother was not married or in a romantic relationship with the often each parent disagreed with their husband/wife/partner over
child’s father (n ⫽ 1,408), and those where neither parent provided issues relating to the child. For FFS, coparenting was based on six
information on couple relationship quality (n ⫽ 3), leaving 2,873 items covering perceived support and discussion of rules, sched-
eligible families. The analytic sample was limited to 2,069 families ules and decisions in raising the child. As for couple supportive-
where information was available on the 9-year-old child’s exter- ness, the average of both parents’ scores (correlations: MCS, r ⫽
nalizing problems from at least one source (parent, teacher, child). .44 p ⬍ .001; FFS, r ⫽ .25, p ⬍ .001) was used in the main
analyses, supplemented by analyses using individual parent per-
Measures
ceptions.
Here, we outline the main measures used in each data set, Negative parenting. For MCS, this used the eight-item
indicating where these drew on items from previously validated Parent–Child Conflict subscale from the Pianta Child–Parent Re-
scales. Internal reliability of measures was generally satisfactory lationship Scale (Pianta, 1992). For FFS, this used a measure of
(␣ ⬎ .70). Where reliability fell slightly below this level in either harsh discipline, concerning frequency of spanking. (A measure of
study, reliability of the equivalent measure in the other study was harsh discipline at ages 3 and 5 (mothers only) was used in
satisfactory. Full details, including sample items, reliability, mean supplementary analyses of MCS, for closer comparison with FFS.)
values, and correlations between measures are provided in the Positive parenting. For MCS, this used the seven-item
online supplemental material. To facilitate comparison of effect Parent–Child Closeness subscale from the Pianta Child–Parent
sizes, all main measures were standardized prior to analysis. relationship scale (Pianta, 1992). For FFS, this used the frequency
Main outcome measure: Middle childhood externalizing of parental involvement in seven activities with the child, such as
problems. These were measured at age 11 (MCS) or age 9 (FFS) reading stories or playing with toys. (Supplementary MCS analy-
using information from parents, teachers and children. Supplemen- ses used a similar age 5 measure of involvement.)
tary analyses in both study samples used additional parent-reported Prior measures of mediators. In FFS (but not MCS), copa-
measures gathered at age 3 (as control variables) and in MCS used renting and parenting were also measured at child aged 1.
parent and teacher reports gathered at age 7 (as supplementary Covariates. These were selected from the literature as poten-
outcome measures). tial confounders of associations between couple supportiveness,
MCS. Parent and teacher reports used the combined conduct coparenting, parenting and externalizing problems. Covariates
and hyperactivity/attentional problems five-item subscales from were measured at 9 months (MCS) or at 1 year (FFS). Parental
the Strengths and Difficulties Questionnaire (Goodman, 1997). sociodemographic characteristics comprised parents’ ages at the
Child reports used items selected for the purposes of this study on birth of the cohort child, whether parents were married, parents’
delinquent behaviors, school engagement, bullying, and anger. ethnicity (MCS: White or ethnic minority; FFS: White non-
FFS. Parent reports (generally from mothers) used items from Hispanic, Black non-Hispanic, Hispanic, and other), and parental
the combined aggressive and rule-breaking subscales of the Child education levels. Household characteristics comprised the number
Behavior Checklist (Achenbach & Rescorla, 2001). Teacher re- of children, whether one or more grandparents lived in the house-
ports used the externalizing problems subscale of the Social Skills hold, whether parents had other children living elsewhere, and
Rating System (Gresham & Elliott, 1990). Child reports used the income in relation to needs (MCS: household income after tax
externalizing subscale of the Self-Description Questionnaire equivalized for household size and composition; FFS: poverty
(Marsh, 1990). ratio, i.e., the ratio of total household income before taxation to the
Primary exposure: Couple supportiveness. Both parents re- U.S. poverty threshold).
ported on the other’s emotional supportiveness when children were Descriptive statistics for sociodemographic covariates are pro-
infants (MCS: 9 months; FFS: 12 months). For MCS, parents vided in the online supplemental material. For each study, there
reported on seven items, including six from the Golombok Rust were statistically significant differences between distributions of
Inventory of Marital State (Rust, Bennun, Crowe, & Golombok, covariates in the analysis sample and a complete sample of couples
1990), concerning the partner’s sensitivity and ability to listen, and at baseline, although these were small in magnitude (averaging less
the respondent’s feelings of closeness, affection, commitment and than a percentage point). For MCS, the analytic sample contained
happiness in the relationship. For FFS, parents reported on five fewer disadvantaged groups (e.g., those with low education or
 COUPLE SUPPORTIVENESS AND CHILD BEHAVIOR PROBLEMS 219

income) than the full couple sample, and fewer ethnic minority stages, corresponding to different elements of the theorized model
groups. For FFS, differences were less consistently associated with (see Figure 1): (1) parenting only, (2) coparenting as an interme-
disadvantage, with the analytic sample containing more Black and diary between couple supportiveness and parenting, (3) as (2), with
unmarried parents. Compared with FFS, the MCS analysis sample an additional direct path from coparenting to externalizing prob-
contained fewer unmarried parents and disadvantaged groups lems (full theorized model). Model fit was assessed at each stage.
(such as parents with no educational qualifications), and was less Indirect effects in the final model were computed as products of
ethnically diverse. All MCS parents were coresiding for children path coefficients. Three Wald tests were used to compare the
aged 1 to 5 years. In FFS, 15% of FFS parents were living apart at magnitude of different sets of indirect pathways (coparenting vs.
child age 1, and this increased to 37% by age 5. parenting, positive vs. negative parenting, and mothers’ vs. fa-
thers’ parenting). Following recommended practice (MacKinnon,
Lockwood, & Williams, 2004), supplementary analyses on non-
Analytic Strategy
imputed data produced bias-corrected bootstrapped estimates of
First, we examined levels and patterns of missing information in indirect effects. (Because of software limitations, the robust max-
the analytic samples. For any variable, on average 4% of MCS and imum likelihood estimator and auxiliary missing information
11% of FFS cases contained incomplete information. This was could not be used in bootstrap models.) To provide more robust
higher for father-reported (averages MCS 4%, FFS, 20%) and tests of the direction of effects, we conducted supplementary
teacher-reported information (MCS 27%, FFS 33%) than for analyses using some repeated measures to assess mediating path-
mother-and child-reported information (MCS 1%, FFS 2% to 3%). ways after allowing for earlier measurement of externalizing prob-
In both samples, incomplete information was more common with lems (both data sets) or prior values of mediators (FFS only).
younger maternal age, maternal ethnic minority, and lower family Supplementary analyses also considered (1) whether indirect ef-
income. To reduce bias, missing information was imputed using fects were similar, depending on whether mothers or fathers pro-
Mplus Version 7.3 (Muthén & Muthén, 1998 –2012). The inclu- vided information on couple supportiveness and coparenting, and
sion in the imputation model of all variables predicting missing- (2) whether alternative measures of negative and positive parent-
ness increases the plausibility of the missing at random assumption ing available in MCS produced similar findings. In these analyses,
(White, Royston, & Wood, 2011). To strengthen prediction of we compared coefficients obtained in different models using z
missing father-reported measures in FFS, the imputation model tests.
included auxiliary variables such as mothers’ reports of fathers’ Analyses were performed on imputed data sets using Mplus
parenting at baseline. As recommended, the number of imputed Version 7.3. For MCS, weights were used to address survey
data sets (70 for each sample) exceeded the percentage of incom- attrition. For FFS, no corresponding weights to address attrition
plete cases (White et al., 2011). Our choice of multiple imputation were available. Available cross-sectional weights for FFS were not
rather than relying on full information maximum likelihood used since oversampling of unmarried and noncoresiding couples
(FIML), another appropriate method available in Mplus to address was considered a virtue for assessing consistency of causal mech-
missing information bias (Graham, 2009), was driven primarily by anisms (though results should not be interpreted as representative
convenience: for each data set it was simpler to conduct a separate of the 20 US cities they were sampled from). Indicator cut-offs
imputation model inclusive of all missingness predictors before applied to assess absolute model fit were ⬍.06 for the root mean
running the various main and supplementary analysis models square error of approximation (RMSEA) and ⬍.08 for the stan-
reported here. As expected, checking our final path models using dardized root mean residual (SRMR; Hu & Bentler, 1999). For
FIML on the original nonimputed analysis samples gave closely path models, comparative fit of models with different sets of
similar findings to those using imputed data. With one exception in indirect pathways was assessed using the Akaike and Bayesian
relation to bootstrapped estimates (see below), which could only information criteria (AIC and BIC, respectively), with smaller
be produced using nonimputed data in Mplus, findings using FIML values indicating better fit. To address clustering of observations in
are not reported here. the MCS sample design and non-normality of measures in both
To address our first hypothesis, we modeled middle childhood analytic samples, we used maximum likelihood estimation with
externalizing problems as a latent construct indicated by parent-, robust standard errors. Throughout, statistical significance was
teacher-, and child-reported information (loadings .5 to .7 in both defined at the p ⬍ .05 level. Estimates for regression and path
data sets). Associations between couple supportiveness and middle models standardized with respect to predictors and outcomes are
childhood externalizing problems were assessed, before and after described, with unstandardized estimates available in the online
adjusting for sociodemographic covariates. Supplementary analy- supplemental material. Because of software constraints, estimates
ses assessed the influence of each parent’s views on partner of indirect effects produced using imputed data are standardized
supportiveness, comparing effects found in separate models using with respect to predictors only.
mothers’ and fathers’ perceptions using z tests of coefficients. In
order to put analyses using continuous measures in a clinical
perspective, supplementary analyses using MCS explored associ- Results
ations between couple supportiveness divided into tertiles (low, In both study samples, couple supportiveness in infancy pre-
medium, high) and severe externalizing problems scored using dicted fewer externalizing problems in middle childhood, with
recommended cut-offs (Goodman, 1997). similar effects in both data sets (standardized betas: MCS ⫽ age
To address our second hypothesis, we used path models to 11, ⫺.15; FFS ⫽ age 9, ⫺.14, both p ⬍ .001, corresponding to the
estimate indirect effects of couple supportiveness via coparenting standard deviation reduction in externalizing problems produced
and parenting mediators. Potential mediators were included in by a one standard deviation increase in couple supportiveness).
220 PARKES, GREEN, AND MITCHELL

These unadjusted effects were slightly reduced after adjusting for a stronger association. A second supplementary analysis examined
sociodemographic covariates (MCS ⫽ age 11, ⫺.13; FFS ⫽ age the association between different tertiles of couple supportiveness
9, ⫺.11, both p ⬍ .001, see Table 1). These effects are based on (high, medium, low) and severe levels of externalizing problems at
the average of both parents’ reports of partner supportivneness. In ages 7 and 11, using MCS data. When compared with couples with
supplementary analyses, z tests comparing effects based solely on high supportiveness, low levels of supportiveness increased the
mothers’ or on fathers’ individual reports of partner supportiveness odds of externalizing problems that could be clinically significant
found stronger associations for mothers’ reports (see Table A in (see Table B in the online supplemental material).
the online supplemental material), although this could reflect Turning to our second hypothesis, we used path models to
shared variance (mothers usually provided parent reports of exter- explore coparenting and parenting as potential mediators of the
nalizing problems). To assess this, we restricted latent constructs effect of couple supportiveness on externalizing problems (parent,
of externalizing problems to teacher and child indicators. In MCS, teacher and child reported) at age 11 (MCS) or age 9 (FFS).
the difference between the effect of mothers’ and fathers’ reports Sociodemographic covariates were allowed to predict mediators
became nonsignificant, while in FFS mothers’ reports still showed and outcome. We explored model fit by including mediators in

Table 1
Regression Models of Associations Between Couple Supportiveness and Middle Childhood
Externalizing Problems (Standardized Estimates)

Unadjusted Adjusted
Measure (reference group) Contrast ␤ SE p ␤ SE p

Millennium Cohort Study (N ⫽ 5,779)

Age 11 externalizing problems
Couple supportiveness –.15 .02 ⬍.001 –.13 .02 ⬍.001
Mother’s age –.07 .02 .005
Father’s age –.06 .02 .014
Number of children .05 .02 .009
Household income –.01 .02 .598
Married at birth –.06 .02 .002
Grandparent in household –.01 .02 .579
Mother has children elsewhere .02 .02 .238
Father has children elsewhere .06 .02 .003
Mother’s education (NVQ 4–5) NVQ Level 2–3 .06 .02 .003
NVQ Level 1 .05 .02 .014
None .11 .03 ⬍.001
Father’s education (NVQ 4–5) NVQ Level 2–3 –.07 .02 .002
NVQ Level 1 –.08 .02 ⬍.001
None .03 .02 .087
Mother’s ethnic group (White) Indian –.02 .02 .145
Pakistani/Bangladeshi –.02 .02 .299
Black .00 .02 .860
Other –.04 .02 .015
Father’s ethnicity different .01 .02 .566

Fragile Families Study (N ⫽ 2,069)

Age 9 externalizing problems
Couple supportiveness –.14 .03 ⬍.001 –.11 .03 ⬍.001
Mother’s age –.12 .05 .015
Father’s age –.01 .05 .890
Number of children –.02 .03 .470
Income to poverty ratio –.07 .03 .044
Married at birth –.04 .04 .279
Grandparent in household .04 .03 .261
Mother has children elsewhere .01 .03 .799
Father has children elsewhere .09 .04 .011
Mother’s education (none) High school –.06 .04 .121
Some college –.12 .04 .002
College –.12 .04 .007
Mother’s ethnic group (White) Black .04 .05 .369
Hispanic –.17 .04 ⬍.001
Other –.02 .03 .421
Father’s ethnicity different –.05 .03 .095
Father more education than mother –.04 .03 .176
Note. NVQ refers to National Vocational Qualifications framework: Levels 4 and 5 correspond to university-
level qualifications (for more details see https://www.gov.uk/what-different-qualification-levels-mean/list-of-
qualification-levels). SE ⫽ standard error.
 COUPLE SUPPORTIVENESS AND CHILD BEHAVIOR PROBLEMS 221

stages as previously described. The final stage (corresponding to effect of coparenting and parenting (60% of the total effect in
the conceptual model in Figure 1) had satisfactory absolute fit MCS, 55% in FFS).
(MCS: ␹2[59] ⫽ 1345.7 p ⬍ .001, RMSEA ⫽ .06, SRMR ⫽ .02; Figure 2 shows statistically significant paths between exposure,
FFS: ␹2[44] ⫽ 157.4 p ⬍ .001, RMSEA ⫽ .04, SRMR ⫽ .01) and mediators and outcome in the final path model for each study
a better fit than earlier stages (based on lowest AIC and BIC sample. Indirect effects are listed in full in Table 2 (with bias-
values; see Table C in the online supplemental material). In both corrected bootstrap confidence intervals produced using nonim-
studies, the direct effect of couple supportiveness on externalizing puted data in the online supplemental material; see Tables D and
problems (i.e., the effect not explained by mediators) was of a E). In MCS, there were significant pathways via all mediators,
similar magnitude (MCS ⫽ ⫺.05, p ⬍ .01; FFS ⫽ ⫺.05, p ⫽ although in FFS the only significant paths were via coparenting
.165). Given that the total effect of couple supportiveness was also and/or mother’s negative parenting. Table 2 also shows the results
similar in both studies, this indicates a similar combined indirect of Wald tests comparing different sets of pathways. In FFS, paths

(a) Millennium Cohort Study

T2 Mother-child
closeness
-.10 (.02)***

T2 Mother-child
.26 (.02)***
.09 (.02)*** -.12 (.02)*** conflict
.08 (.02)***
-.20 (.01)***
.27 (.01)***
T1 Couple T2 Coparenng -.10 (.02)***
T3 Child
supporveness .03 (.02)* externalizing
.13 (.02)*** -.22 (.02)*** -.05 (.02)** behavior
T2 Father-child
-.13 (.02)*** closeness
-.05 (.02)**

.04 (.02)**

T2 Father-child
conflict

(b) Fragile Families Study

T2 Mother’s
involvement

T2 Mother’s harsh
.17 (.02)*** -.10 (.03)*** discipline .17 (.04)***
-.07 (.03)*

T1 Couple .33 (0.03)*** T2 -.16 (.04)***

supporveness T3 Child
Coparenng
externalizing
.12 (.03)*** .15 (.05)*** behavior

-.07 (.03)* T2 Father’s

involvement

T2 Father’s harsh
discipline

Figure 2. Final path models of associations between couple supportiveness in infancy and middle childhood
externalizing problems. (a) Millennium Cohort Study (MCS). (b) Fragile Families Study (FFS). Model fit: (a)
MCS: ␹2(59) ⫽ 1345.7 p ⬍ .001, root mean square error of approximation (RMSEA) ⫽ .06, standardized root
mean residual (SRMR) ⫽ .02; (b) FFS: ␹2(44) ⫽ 157.4 p ⬍ .001, RMSEA ⫽ .04, SRMR ⫽ .01. Child ages
at T1 ⫽ 9 months (MCS), 1 year (FFS), T2 ⫽ 3/5 years (MCS), 3 years (FFS), T3 ⫽ 11 years (MCS), 9
years (FFS). Models adjusted for parents’ age, ethnicity, education, marital status, number of children,
nonresident children, resident grandparent, and household income. Correlations modeled between T2
parenting measures, and all nonsignificant associations between constructs, have been omitted. Figures
show standardized coefficients with standard errors in parentheses. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01. ⴱⴱⴱ p ⬍ .001.
222 PARKES, GREEN, AND MITCHELL

Table 2
Indirect Effects From Couple Supportiveness to Children’s Externalizing Problems

Mediator(s) Effect SE p Mediator(s) Effect SE p

Millennium Cohort Study Fragile Families Study

(child age 11) (child age 9)
Coparenting only –.029 .005 ⬍.001 Coparenting only –.054 .015 ⬍.001
Coparenting and mother–child closeness –.002 .001 ⬍.001 Coparenting and mother’s involvement .000 .001 .978
Coparenting and mother–child conflict –.015 .002 ⬍.001 Coparenting and mother’s harsh discipline –.004 .002 .028
Coparenting and father–child closeness .000 .000 .108 Coparenting and father’s involvement .001 .002 .475
Coparenting and father–child conflict –.003 .001 .010 Coparenting and father’s harsh discipline .001 .001 .493
Mother–child closeness only –.010 .002 ⬍.001 Mother’s involvement only .000 .006 .980
Mother–child conflict only –.035 .005 ⬍.001 Mother’s harsh discipline only –.016 .006 .005
Father–child closeness only –.007 .003 .007 Father’s involvement only .004 .005 .454
Father–child conflict only –.006 .002 .012 Father’s harsh discipline only –.003 .003 .443
All via mother’s parenting –.063 .006 ⬍.001 All via mother’s parenting –.020 .008 .015
All via father’s parenting –.016 .004 ⬍.001 All via father’s parenting .003 .007 .654
All via coparenting (⫹/⫺parenting) –.052 .006 ⬍.001 All via coparenting (⫹/⫺parenting) –.055 .015 ⬍.001
All via parenting only –.058 .007 ⬍.001 All via parenting only –.016 .009 .089
All via positive parenting –.019 .004 ⬍.001 All via positive parenting .005 .009 .574
All via negative parenting –.060 .006 ⬍.001 All via negative parenting –.022 .006 ⬍.001

Comparison of pathways Wald test df p Comparison of pathways Wald test df p

Mother’s vs. father’s parenting 68.73 1 ⬍.001 Mother’s vs. father’s parenting 4.00 1 .045
Coparenting (⫹/⫺parenting) vs. parenting only .92 1 .338 Coparenting (⫹/⫺parenting) vs. parenting only 4.80 1 .029
Positive vs. negative parenting 32.29 1 ⬍.001 Positive vs. negative parenting 5.75 1 .017
Note. These represent indirect effects in the final path models. Effects are standardized with respect to predictors only. SE ⫽ standard error.

involving coparenting were stronger than those involving parent- child conflict. As in the main model (see Table 2), significant
ing alone, although in MCS there was no difference. In both data indirect pathways were found via positive and negative parenting,
sets, pathways via negative parenting were stronger than via pos- indicating broad equivalence of alternative study parenting mea-
itive parenting, and paths were stronger for mothers’ than for sures (see Table I in the online supplemental material). Nonethe-
fathers’ parenting. less, z tests showed that paths via alternative MCS parenting
In both study samples, further models allowed for parent- measures were reduced in size. This suggests that in the MCS data
reported externalizing problems at child age 3. This provided a test set, involvement and harsh discipline measures (equivalent to the
of whether coparenting and parenting mediated effects of couple FFS parenting measures) did not provide such a sensitive test of
supportiveness on changes in externalizing problems from pre- the effect of couple supportiveness on parenting, as compared with
school age to age 11 (MCS) or age 9 (FFS). After allowing for measures of parent– child closeness and conflict.
concurrent externalizing problems, all paths via coparenting and/or
mother’s negative parenting remained significant in both data sets
Discussion
(see Table F in the online supplemental material). In FFS, a further
model allowing for coparenting and parenting at child age 1 This study found that more supportive couple relationships in
provided a more robust test of whether couple supportiveness at early childhood reduced externalizing problems in middle child-
age 1 produced a change in mediators at age 3. In this model, paths hood (including reduced risk of clinically significant problems).
via coparenting alone and mother’s negative parenting alone both Much of the effect of couple supportiveness was explicable in
remained statistically significant (see Figure A and Table G in the terms of coparenting and parenting behavior during preschool
online supplemental material). Two sets of supplementary analyses years. This is the first longitudinal study finding spillover path-
assessed the sensitivity of indirect effects to the source of infor- ways to middle childhood adjustment from couple supportiveness
mation, and to the different parenting measures used in the two in infancy via both coparenting and parenting. It extends our
studies. Results showed a high degree of consistency. In relation to understanding of spillover involving coparenting, from previous
source of information, there were significant indirect pathways cross-sectional work (O’Leary & Vidair, 2005; Stroud et al.,
involving coparenting (with or without parenting), regardless of 2015). It confirms the relevance of parenting spillover to effects of
whether fathers or mothers supplied information on couple sup- positive relationship qualities, previously documented for couple
portiveness and coparenting (see Table H in the online supplemen- conflict (Gerard et al., 2006; Lindsey et al., 2009; Rhoades et al.,
tal material). In the MCS data set (but not the FFS data set), z tests 2011; Schoppe-Sullivan et al., 2007; Stover et al., 2016). Strengths
showed combined indirect pathways were larger in magnitude of the study include the use of information from both parents to
when mothers’ perceptions of supportiveness and coparenting assess relationship quality, parenting and coparenting; outcome
were used. Supplementary MCS analyses substituted measures of information from parents, teachers and children to reduce effects
parental involvement (age 5) for parent– child closeness (age 3), of shared method variance; and allowance for confounding. A
and mother’s use of harsh discipline (ages 3 and 5) for mother– broadly consistent pattern of results across two population-based
 COUPLE SUPPORTIVENESS AND CHILD BEHAVIOR PROBLEMS 223

samples from different contexts, with the U.S. sample incorporat- with stronger effects for fathers’ than mothers’ hostile parenting
ing more unmarried and separated parents than previously studied, found in two studies (Rhoades et al., 2011; Stover et al., 2016) and
permits greater generalizability of study findings. another study finding no difference (Stroud et al., 2015); but these
The total effect size of couple supportiveness was similar in studies used couple-based convenience samples, which may over-
magnitude across the two study samples, although smaller than represent families where fathers are more involved in parenting.
effects typically found elsewhere (Buehler et al., 1997; Reid & Our findings from population-based samples may reflect a more
Crisafulli, 1990). This could be because our meaure of couple typical picture of the mother as the child’s main caregiver.
supportiveness focused on more positive aspects of the couple Although this study lends strong empirical support to spillover
relationship rather than on marital conflict or physical violence: mechanisms, including both parenting and coparenting, the medi-
our effect sizes are closer to those of a meta-analysis that included ators explored here did not fully explain associations between
measures of general marital distress as well as conflict (.16; Reid couple relationship quality and child externalizing problems.
& Crisafulli, 1990) than to effects from a meta-analysis focusing Moreover, these observational associations only indicate causal
specifically on conflict (.32; Buehler et al., 1997). Previous studies effects under the assumption of no unmeasured confounding. De-
have also relied heavily on cross-sectional data, compared with our spite the use of several strategies to strengthen causal inference,
8- to 10-year follow-up period. unmeasured confounding remains a possibility, for example, from
Our findings in relation to mediators strengthen the limited parental temperament. Last, although analyses using repeated mea-
existing evidence for spillover effects of the couple relationship sures supported the idea of a directional effect from the couple
onto two types of family process linked with externalizing prob- relationship to child adjustment via mediators, our study design
lems during middle childhood (O’Leary & Vidair, 2005; Stroud et does not allow for the likely bidirectionality of associations be-
al., 2015). Pathways involving coparenting were more important tween measures over time indicated by other studies (e.g., Gold-
(U.S.) or just as important (U.K.) as those mediated by parenting berg & Carlson, 2014). Further work is needed to address these
alone. The greater importance of coparenting in the U.S. study may issues, using repeated measurements to construct fixed effects and
reflect superior measurement, set against better measures of par- cross-lagged models, or using marginal structural models to adjust
enting in the U.K. study. Overall, our study confirms previous for time-varying confounding.
cross-sectional work viewing coparenting as an intermediary be-
Limitations in addition to those already noted include reliance
tween the couple relationship and parenting (O’Leary & Vidair,
on parents for sensitive information subject to social desirability
2005). It also supports cross-sectional work indicating that copa-
and other biases and use of mediator measures that may not
renting transmits additional effects of the couple relationship on
adequately capture underlying constructs. Notably, coparenting
children’s behavior, over and above those conveyed via parenting
measures did not fully encompass the multidimensional nature of
(Stroud et al., 2015). For young children, triadic coparenting
coparenting theorized elsewhere (Feinberg, 2003) and were not
disagreements may be more salient than similar, more direct,
previously validated in either data set. The study samples differed
effects of dyadic couple interactions: The child is likely to be
in measures used for various constructs and the age of children at
present and sensitive to matters concerning him/herself (Feinberg,
the final outcome, making it is difficult to determine what causes
2003; Talbot & McHale, 2004).
between-study differences (differences in measurement, sampling,
In both study samples, spillover pathways found via negative
parenting echo findings from other studies of families with young context, etc.). Measurement limitations of this type are common
children, via parental harsh or inconsistent discipline, hostility, when comparing across large multipurpose population samples,
overreactivity, and psychological control (Gerard et al., 2006; though this is counterbalanced by already-noted strengths includ-
O’Leary & Vidair, 2005; Rhoades et al., 2011; Schoppe-Sullivan ing generalizability and consistency across different contexts.
et al., 2007; Stover et al., 2016). In the U.K. sample, there were However, the FFS sample is not fully representative of the United
also smaller pathways involving positive parenting. This, too, States, so analyses should not be interpreted as a straight U.S.–
supports other work on families with young children finding U.K. comparison.
pathways via positive parenting, including warmth, involvement, Overall, our study supports a broad model of spillover effects on
emotional reciprocity and attachment security (Gerard et al., 2006; child adjustment, operating consistently across different settings.
Lindsey et al., 2009; Schoppe-Sullivan et al., 2007). The stronger Future research using population-based samples would benefit
effect of negative parenting compared with positive parenting we from more detailed measures of the couple relationship, including
found overall might reflect our available measures, but is striking both positive and negative dimensions, and a fuller assessment of
given that our primary exposure of couple supportiveness did not family processes. It is reasonable to assume that our coparenting,
specifically measure couple conflict. It chimes with a population and parenting measures gathered at preschool and school transition
study of marital conflict, which also found larger paths to exter- reflected earlier as well as concurrent processes (Dallaire & Wein-
nalizing problems via negative parenting (Gerard et al., 2006). raub, 2005; Laxman et al., 2013), but future work should use
Stronger pathways were found for mothers’ than for fathers’ earlier measurements to better assess impacts on infant and toddler
parenting in both samples. This might reflect higher levels of development. Our own study suggests a particular need to explore
missing information for fathers, and/or shared method variance direct effects of coparenting in more detail. Our measures did not
inflating associations between maternal parenting and mother- capture whether the child was present during coparenting, but
reported child externalizing problems. Nonetheless, we took pre- future work could examine young children’s psychological re-
cautions to guard against both these risks, in strengthening impu- sponses to covert versus overt coparenting tensions, explore pos-
tation of missing father information with auxiliary information and sible moderators and investigate additional aspects of children’s
using multiinformant outcome measures. Our findings contrast adjustment.
224 PARKES, GREEN, AND MITCHELL

Our study offers greater clarity on the contribution of coparent- Science and Practice, 3, 95–131. http://dx.doi.org/10.1207/
ing to causal mechanisms linking the overall couple relationship in S15327922PAR0302_01
the important vulnerable period of early childhood to later child Fincham, F. D., & Hall, J. H. (2005). Parenting and the marital relationship.
behavior. The findings lend strong support to parenting interven- In T. Luster & L. Okagaki (Eds.), Parenting: An ecological perspective
tions that include coparenting as well as dyadic couple relationship (2nd ed., pp. 205–234). Mahwah, NJ: Lawrence Erlbaum.
Gerard, J. M., Krishnakumar, A., & Buehler, C. (2006). Marital conflict,
and individual parenting skills. In recent years there has been a
parent-child relations, and youth maladjustment—A longitudinal inves-
surge of interest in interventions specifically designed to help
tigation of spillover effects. Journal of Family Issues, 27, 951–975.
couples in the early stages of parenthood manage the stresses and http://dx.doi.org/10.1177/0192513X05286020
strains involved in child rearing, by developing their coparenting Goldberg, J. S., & Carlson, M. J. (2014). Parents’ relationship quality and
skills (Pruett, Pruett, Cowan, & Cowan, 2017). The transition to children’s behavior in stable married and cohabiting families. Journal of
parenthood may be an ideal opportunity to intervene to help Marriage and the Family, 76, 762–777. http://dx.doi.org/10.1111/jomf
families, being a period when couples are receptive to advice and .12120
are less likely to have well-established patterns of coparenting Goodman, R. (1997). The Strengths and Difficulties Questionnaire: A
(Feinberg, 2002). Interventions to facilitate coparenting, though research note. Journal of Child Psychology and Psychiatry, and Allied
presenting many challenges for couples with unstable relationships Disciplines, 38, 581–586. http://dx.doi.org/10.1111/j.1469-7610.1997
(McHale, Waller, & Pearson, 2012), also may be more attractive to .tb01545.x
parents than relationship counseling (Doss, Cicila, Hsueh, Morri- Graham, J. W. (2009). Missing data analysis: Making it work in the real
son, & Carhart, 2014). world. Annual Review of Psychology, 60, 549 –576. http://dx.doi.org/10
.1146/annurev.psych.58.110405.085530
Gresham, F. M., & Elliott, S. N. (1990). Social Skills Rating System. Circle
References Pines, MN: American Guidance Service.
Holland, A. S., & McElwain, N. L. (2013). Maternal and paternal percep-
Achenbach, T. M., & Rescorla, L. A. (2001). Manual for the ASEBA tions of coparenting as a link between marital quality and the parent-
school-age forms & profiles. Burlington, VT: University of Vermont, toddler relationship. Journal of Family Psychology, 27, 117–126. http://
Research Center for Children, Youth & Families. dx.doi.org/10.1037/a0031427
Belsky, J. (1984). The determinants of parenting: A process model. Child Howard, K. S., & Brooks-Gunn, J. (2009). Relationship Supportiveness
Development, 55, 83–96. http://dx.doi.org/10.2307/1129836 during the transition to parenting among married and unmarried parents.
Buehler, C., Anthony, C., Krishnakumar, A., Stone, G., Gerard, J., & Parenting, 9, 123–142. http://dx.doi.org/10.1080/15295190802656828
Pemberton, S. (1997). Interparental conflict and youth problem behav- Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in
iors: A meta-analysis. Journal of Child and Family Studies, 6, 233–247. covariance structure analysis: Conventional criteria versus new al-
http://dx.doi.org/10.1023/A:1025006909538 ternatives. Structural Equation Modeling, 6, 1–55. http://dx.doi.org/
Chen, M., & Johnston, C. (2012). Interparent childrearing disagreement, 10.1080/10705519909540118
but not dissimilarity, predicts child problems after controlling for par- Karreman, A., van Tuijl, C., van Aken, M. A. G., & Dekovic, M. (2008).
enting effectiveness. Journal of Clinical Child and Adolescent Psychol- Parenting, coparenting, and effortful control in preschoolers. Journal of
ogy, 41, 189 –201. http://dx.doi.org/10.1080/15374416.2012.651997
Family Psychology, 22, 30 – 40. http://dx.doi.org/10.1037/0893-3200.22
Cummings, E. M., Merrilees, C. E., & Ward George, M. (2010). Fathers,
.1.30
marriages and families: Revisiting and updating the framework for
Krishnakumar, A., & Buehler, C. (2000). Interparental conflict and par-
fathering in context. In M. E. Lamb (Ed.), The role of the father (5th ed.,
enting behaviors: A meta-analytic review. Family Relations, 49, 25– 44.
pp. 154 –176). Hoboken, NJ: Wiley.
http://dx.doi.org/10.1111/j.1741-3729.2000.00025.x
Dallaire, D. H., & Weinraub, M. (2005). The stability of parenting behav-
Laxman, D. J., Jessee, A., Mangelsdorf, S. C., Rossmiller-Giesing, W.,
iors over the first 6 years of life. Early Childhood Research Quarterly,
Brown, G. L., & Schoppe-Sullivan, S. J. (2013). Stability and anteced-
20, 201–219. http://dx.doi.org/10.1016/j.ecresq.2005.04.008
ents of coparenting quality: The role of parent personality and child
Doss, B. D., Cicila, L. N., Hsueh, A. C., Morrison, K. R., & Carhart, K.
temperament. Infant Behavior and Development, 36, 210 –222. http://dx
(2014). A randomized controlled trial of brief coparenting and relation-
.doi.org/10.1016/j.infbeh.2013.01.001
ship interventions during the transition to parenthood. Journal of Family
Psychology, 28, 483– 494. http://dx.doi.org/10.1037/a0037311 Lindsey, E. W., Caldera, Y. M., & Tankersley, L. (2009). Marital conflict
Erel, O., & Burman, B. (1995). Interrelatedness of marital relations and and the quality of young children’s peer play behavior: The mediating
parent-child relations: A meta-analytic review. Psychological Bulletin, and moderating role of parent– child emotional reciprocity and attach-
118, 108 –132. http://dx.doi.org/10.1037/0033-2909.118.1.108 ment security. Journal of Family Psychology, 23, 130 –145. http://dx.doi
Fay-Stammbach, T., Hawes, D. J., & Meredith, P. (2014). Parenting .org/10.1037/a0014972
influences on executive function in early childhood: A review. Child Mackinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence
Development Perspectives, 8, 258 –264. http://dx.doi.org/10.1111/cdep limits for the indirect effect: Distribution of the product and resampling
.12095 methods. Multivariate Behavioral Research, 39, 99 –99. http://dx.doi
Fearon, R. P., Bakermans-Kranenburg, M. J., van Ijzendoorn, M. H., .org/10.1207/s15327906mbr3901_4
Lapsley, A.-M., & Roisman, G. I. (2010). The significance of insecure Marsh, H. W. (1990). Self-Description Questionnaire Manual. Campbell-
attachment and disorganization in the development of children’s exter- town, New South Wales, Australia: University of Western Sydney,
nalizing behavior: A meta-analytic study. Child Development, 81, 435– Macarthur.
456. http://dx.doi.org/10.1111/j.1467-8624.2009.01405.x McHale, J., Waller, M. R., & Pearson, J. (2012). Coparenting interventions
Feinberg, M. E. (2002). Coparenting and the transition to parenthood: A for fragile families: What do we know and where do we need to go next?
framework for prevention. Clinical Child and Family Psychology Re- Family Process, 51, 284 –306. http://dx.doi.org/10.1111/j.1545-5300
view, 5, 173–195. http://dx.doi.org/10.1023/A:1019695015110 .2012.01402.x
Feinberg, M. E. (2003). The internal structure and ecological context of Moss, E., Cyr, C., Bureau, J. F., Tarabulsy, G. M., & Dubois-Comtois, K.
coparenting: A framework for research and intervention. Parenting: (2005). Stability of attachment during the preschool period. Develop-
 COUPLE SUPPORTIVENESS AND CHILD BEHAVIOR PROBLEMS 225

mental Psychology, 41, 773–783. http://dx.doi.org/10.1037/0012-1649 Reichman, N. E., Teitler, J. O., Garfinkel, I., & McLanahan, S. S. (2001).
.41.5.773 Fragile families: Sample and design. Children and Youth Services Re-
Murphy, S. E., Jacobvitz, D. B., & Hazen, N. L. (2016). What’s so bad view, 23, 303–326. http://dx.doi.org/10.1016/S0190-7409(01)00141-4
about competitive coparenting? Family-level predictors of children’s Reid, W. J., & Crisafulli, A. (1990). Marital discord and child behavior
externalizing symptoms. Journal of Child and Family Studies, 25, problems: A meta-analysis. Journal of Abnormal Child Psychology, 18,
1684 –1690. http://dx.doi.org/10.1007/s10826-015-0321-5 105–117. http://dx.doi.org/10.1007/BF00919459
Muthén, L. K., & Muthén, B. O. (1998 –2012). Mplus user’s guide (7th Rhoades, K. A., Leve, L. D., Harold, G. T., Neiderhiser, J. M., Shaw, D. S.,
ed.). Los Angeles, CA: Authors. & Reiss, D. (2011). Longitudinal pathways from marital hostility to
Odgers, C. L., Moffitt, T. E., Broadbent, J. M., Dickson, N., Hancox, R. J., child anger during toddlerhood: Genetic susceptibility and indirect ef-
Harrington, H., . . . Caspi, A. (2008). Female and male antisocial fects via harsh parenting. Journal of Family Psychology, 25, 282–291.
trajectories: From childhood origins to adult outcomes. Development http://dx.doi.org/10.1037/a0022886
and Psychopathology, 20, 673–716. http://dx.doi.org/10.1017/ Rust, J., Bennun, I., Crowe, M., & Golombok, S. (1990). The GRIMS. A
S0954579408000333 psychometric instrument for the assessment of marital discord. Journal
O’Leary, S. G., & Vidair, H. B. (2005). Marital adjustment, child-rearing of Family Therapy, 12, 45–57. http://dx.doi.org/10.1046/j.1990.00369.x
disagreements, and overreactive parenting: Predicting child behavior Schoemaker, K., Mulder, H., Deković, M., & Matthys, W. (2013). Exec-
problems. Journal of Family Psychology, 19, 208 –216. http://dx.doi.org/ utive functions in preschool children with externalizing behavior prob-
10.1037/0893-3200.19.2.208 lems: A meta-analysis. Journal of Abnormal Child Psychology, 41,
Pedro, M. F., Ribeiro, T., & Shelton, K. H. (2012). Marital satisfaction and 457– 471. http://dx.doi.org/10.1007/s10802-012-9684-x
partners’ parenting practices: The mediating role of coparenting behav- Schoppe-Sullivan, S. J., Schermerhorn, A. C., & Cummings, E. M. (2007).
ior. Journal of Family Psychology, 26, 509 –522. http://dx.doi.org/10
Marital conflict and children’s adjustment: Evaluation of the parenting
.1037/a0029121
process model. Journal of Marriage and the Family, 69, 1118 –1134.
Pendry, P., & Adam, E. K. (2013). Child-related interparental conflict in
http://dx.doi.org/10.1111/j.1741-3737.2007.00436.x
infancy predicts child cognitive functioning in a nationally representa-
Stover, C. S., Zhou, Y., Kiselica, A., Leve, L. D., Neiderhiser, J. M., Shaw,
tive sample. Journal of Child and Family Studies, 22, 502–515. http://
D. S., . . . Reiss, D. (2016). Marital hostility, hostile parenting, and child
dx.doi.org/10.1007/s10826-012-9603-3
aggression: Associations from toddlerhood to school age. Journal of the
Pérez, F., Moessner, M., & Santelices, M. P. (2017). Beyond the dyad: The
American Academy of Child & Adolescent Psychiatry, 55, 235–242.
relationship between preschoolers’ attachment representations and fam-
http://dx.doi.org/10.1016/j.jaac.2015.12.008
ily triadic interactions. Infant Mental Health Journal, 38, 198 –209.
Stroud, C. B., Meyers, K. M., Wilson, S., & Durbin, C. E. (2015). Marital
http://dx.doi.org/10.1002/imhj.21633
Petersen, I. T., Bates, J. E., Dodge, K. A., Lansford, J. E., & Pettit, G. S. quality spillover and young children’s adjustment: Evidence for dyadic
(2015). Describing and predicting developmental profiles of externaliz- and triadic parenting as mechanisms. Journal of Clinical Child and
ing problems from childhood to adulthood. Development and Psycho- Adolescent Psychology, 44, 800 – 813. http://dx.doi.org/10.1080/
pathology, 27, 791– 818. http://dx.doi.org/10.1017/S0954579414000789 15374416.2014.900720
Pianta, R. C. (1992). Child-Parent Relationship Scale. (Unpublished mea- Talbot, J. A., & McHale, J. P. (2004). Individual parental adjustment
sure). Charlottesville, VA: Curry School of Education, University of moderates the relationship between marital and coparenting quality.
Virginia. Journal of Adult Development, 11, 191–205. http://dx.doi.org/10.1023/
Pinquart, M. (2017). Associations of parenting dimensions and styles with B:JADE.0000035627.26870.f8
externalizing problems of children and adolescents: An updated meta- Teubert, D., & Pinquart, M. (2010). The association between coparenting
analysis. Developmental Psychology, 53, 873–932. http://dx.doi.org/10 and child adjustment: A meta-analysis. Parenting: Science and Practice,
.1037/dev0000295 10, 286 –307. http://dx.doi.org/10.1080/15295192.2010.492040
Plewis, I., Calderwood, L., Hawkes, D., Hughes, G., & Joshi, H. (2007). White, I. R., Royston, P., & Wood, A. M. (2011). Multiple imputation
The Millennium Cohort Study. (Tech. report). London, UK: Centre for using chained equations: Issues and guidance for practice. Statistics in
Longitudinal Studies, Bedford Group for Lifecourse and Statistical Stud- Medicine, 30, 377–399. http://dx.doi.org/10.1002/sim.4067
ies, Institute of Education, University of London.
Pruett, M. K., Pruett, K., Cowan, C. P., & Cowan, P. A. (2017). Enhancing
father involvement in low-income families: A couples group approach to Received March 26, 2018
preventive intervention. Child Development, 88, 398 – 407. http://dx.doi Revision received October 25, 2018
.org/10.1111/cdev.12744 Accepted October 26, 2018 䡲